Fix sparse.dot gradient for 1d dense operands

[jarotter]

Description

pytensor.sparse.dot supports a 1d dense operand — Dot.make_node accepts it, infer_shape handles it, and the forward pass computes correctly (covered by the existing test_csr_dense/test_csc_dense tests). But Dot.pullback assumed both operands were always 2d, so differentiating anything containing sparse.dot(X, beta) (or sparse.dot(beta, X)) with a 1d beta raised inside pytensor.grad itself, before any compilation.

Depending on whether the sparse operand's static shape was known at graph-construction time, this surfaced as one of two different errors:

ValueError: Incompatible shared dimension for dot product: (1, 20), (5, 1)

or

TypeError: The dense dot product is only supported for dense types

Both are the same underlying defect, just diverging on which branch of pullback hits a shape check first.

Fix: promote the 1d operand, and the incoming gradient gz, to a 2d row/column before building the gradient, reuse the existing matrix-product gradient formulas unchanged, then squeeze the added axis back out of the result — mirroring the promotion dense_dot already does for the dense Dot op. Behavior for the already-working 2d/2d case is untouched (verified with dprint tcase is byte-identical to before).

Testing

Added 8 tests to TestDots in tests/sparse/test_math.py:

• test_dot_vector_operand_grad — the core fix, both operand ord
• test_dot_vector_operand_grad_length_one — broadcastable lengt
• test_dot_vector_operand_grad_shape_regimes — both failure-modlved shape)
• test_dot_vector_operand_grad_squared_error_loss — end-to-end rted pattern
• test_dot_matrix_sparse_operand_grad — non-regression: grad w.e already-working 2d case, previously untested

Related Issue

• Closes BUG: grad of sparse.dot fails when the operand is a 1d vector #2282

Checklist

• Checked that the pre-commit linting/style checks pass
• Included tests that prove the fix is effective or that the new feature works
• Added necessary documentation (docstrings and/or example notebooks)
• If you are a pro: each commit corresponds to a [relevant logical change]

Type of change

• New feature / enhancement
• Bug fix
• Documentation
• Maintenance
• Other (please specify):